CONFIGURABLE PRODUCTS − LESSONS LEARNED FROM THE

advertisement
CONFIGURABLE PRODUCTS −
LESSONS LEARNED FROM THE FINNISH INDUSTRY
Juha Tiihonen, Timo Soininen, Tomi Männistö, and Reijo Sulonen
Helsinki University of Technology
TAI Research Centre, Product Data Management Group
P.O. Box 9555, FIN-02015 HUT, Finland
Email: Juha.Tiihonen@hut.fi
ABSTRACT
In this paper we present experiences on configurable products gathered in co-operation with Finnish companies. The
reasons that made configurable products important include the ability to efficiently fulfil a wide range of customer
requirements, shorter lead times in the sales-delivery process and increased control of production. We discuss the
product development and sales-delivery processes of configurable products and their practical problems. These
included partially configurable products, poor quality of configuration models and lack of clearly defined product
policy. Companies had migrated to configurable products from one-of-a-kind products and fixed products (i.e.
products manufactured repetitively according to a fixed specification). The direction of migration affects the
benefits, problems and necessary changes. The requirements of the sales-delivery process and the product must be
understood to gain full benefit from a product configurator. In addition, the process must usually be re-engineered
and products designed for configurability. Companies often neglect these issues and the long-term management of
configuration knowledge. This leads to major difficulties in configurator projects.
KEYWORDS
Product configuration; configurators; configuration experiences; survey
1.
Introduction and definition of configurable products
Configurable products are important for industries that offer products adapted according to customer requirements.
Information systems that support the customer specific adaptation, i.e. product configurators, are an interesting
subject for research (see e.g.[1]). However, in our view product configuration is a much broader research area. The
success of applying a configurator depends on the related business issues, products, processes, organizations and
practices. Only through understanding these issues can configurators and configurable products be applied to their
full potential. Such understanding is also valuable for developing configurators.
This paper reports experiences on configurable products gained in co-operation with Finnish companies. The results
are largely based on an in-depth survey of ten companies[2]. Some experiences are derived from close co-operation
with half a dozen other companies that use or plan to use product configurators. The companies were interested in
developing their processes, products, information systems and knowledge management related to product configuration. They were neither a random sample nor a collection of state-of-the-art companies.
Our definition of the basic properties of configurable products is as follows:
• The product has been pre-designed to meet a given range of different customer requirements.
• Each delivered product individual is adapted to the needs of a customer.
• Each product individual is specified as an arrangement of pre-designed components. Thus, there is no need to
design new components as a part of the sales-delivery process.
• The product has a pre-designed architecture.
• No creative or innovative design is needed as a part of the sales-delivery process. Rather, a product individual
can be specified in a routine manner.
This definition coincided with how the companies perceived configurable products.
Number of variants
Number of
modules
Number of parts
in modules
2.
Reasons for operating with configurable products
We asked representatives of the companies in the survey to explain why they
design and manufacture configurable products. We first list the reasons and
then analyze them. The reasons mentioned most often were:
• ability to fulfil a wide range of customer requirements (7 companies),
• shorter lead times in the sales-delivery process (5),
• increased control of production (4),
• reduction in customer-specific design (4),
• efficient way to offer a broad product line (3), and
• improved quality (3).
Other reasons mentioned included easier selling, less capital tied up to work in
progress and stock, and reduction of errors.
Figure 1. Controllability of
configurable products
The ability to efficiently fulfil a wide range of customer requirements is the
main reason for designing and manufacturing configurable products. This is
vital in today’s highly competitive global markets. Configurable products usually have a modular product architecture, where each module implements completely one or a few functions, and where interactions between modules are
well defined[3]. This allows a broad product line as combinations of relatively few pre-designed modules.
Assemble-to-order production of configurable products is possible in many cases. This reduces lead times compared
to one-of-a-kind products, which are engineered to order. In addition, the time and effort required for customer
specific design can be practically eliminated. This allows companies to free engineering resources from customer
specific design. Many companies wanted to use these resources for product development. This can be a significant
factor, especially on markets with time-based competition.
As configurable products consist of relatively few pre-designed and independent modules, control of production is
easier compared to one-of-a-kind products. There are fewer items to forecast and manage compared to an equivalent
range of fixed products (Figure 1). Each module can have a fairly high volume making forecasting easier and
manufacturing more economical. Fewer items are needed in the stock and less capital is tied up to work in progress
and stock. Note that configurable products can be controlled effectively at module level. There is little justification
for forecasting individual product variants, although some of the companies tried to do this with mediocre success.
Configurable products are easier to sell than one-of-a-kind products in the sense that the possibilities of the product
have been decided in advance. This gives a more concrete point of reference in communicating what is possible to
the customer. Pricing is usually much easier than for one-of-a-kind products. It is also easier to prepare supporting
material for the different phases of the sales-delivery process.
Surprisingly, reduction of errors was explicitly mentioned only once by the companies surveyed. This reason may
have been included as part of improved quality. In-depth discussions with some companies implied that reducing the
number of errors is an important factor.
3.
Processes related to configurable products
In this section we discuss the two main processes related to products: the product development process and the salesdelivery process. The product development process of a configurable product produces, among other deliverables, a
configuration model. A configuration model contains all the information on the possibilities of adapting the product
to customer needs. It consists of available components, rules on correct component combinations, and rules on how
to achieve the desired functions for a customer. The configuration model is used repeatedly in the configuration
process (Figure 2) that is a part of the sales-delivery process, to produce configurations, i.e. descriptions of the
product individuals to be delivered. In the configuration process product individuals are configured, i.e. adapted to
meet given customer requirements, on the basis of the configuration model and the customer requirements. These
adaptation activities are referred to as configuration tasks. We call a person that configures a product a configurer.
The product development and sales-delivery processes of configurable products are separate. In this regard a
configurable product resembles a fixed product and differs from a one-of-a-kind product. However, configurable
products differ from fixed products since the latter are not adapted for each customer. The product development
process of a one-of-a-kind product can be seen as a part of the sales-delivery process. It typically requires creative or
innovative design that results in a specification of a single product individual. Creative or innovative design makes it
possible to design new component types and to use an arbitrary product architecture. In principle, it becomes
Customer
Requirements
Sales
configuration
process
Customer spesific
product individual
Configuration
model 1
Configuration 1
Specification
Configuration
model 2
Engineering
configuration
process
Configuration 2
Description of
product individual
Component
orders
Product
individual
Assembly
and/or
installation
Components
Component
manufacturing
and/or stock
Figure 2. Sales-delivery process of
configurable products
possible to satisfy any customer requirements. Some parts of the
specification may be re-used in more or less ad-hoc manner in other
products. A configurable product significantly enhances the re-use of
design knowledge compared to a similar one-of-a-kind product.
Product development process
Identifying the market segments and their specific requirements are
important in developing a configurable product. A company must
choose the segments to which the product and each of its variation
possibilities are targeted. It is critical to match the product architecture and variation possibilities to typical combinations of requirements. A mismatch makes the product hard to sell because suitable
alternatives are not available by configuring the product.
A preliminary configuration model should be defined early in the
product development process to define the variation possibilities of
the product. This is an important part of the conceptual and systemlevel design of the product. It is vital for the configuration process to
document the configuration model systematically. There may be
different configuration models of the same product. These are
typically meant to support different phases of the configuration
process, for example sales and engineering configuration tasks.
A product expert with clear understanding of the product should enter the configuration model in a configurator. If a
knowledge engineer enters the configuration model, he or she should not need to re-invent the configuration model
or to gather the required information from multiple sources. Product policy and commercial and technical restrictions
of the product are typically clearly beyond the expertise and authority of knowledge engineers.
As discussed earlier, modular product architecture is almost a necessary characteristic of a configurable product.
Note that some parametric products can be considered configurable although they are not necessarily modular. In our
view, the main distinguishing characteristics between a non-configurable and a configurable modular product are that
the latter has a systematically documented configuration model and a systematic configuration process.
Most companies offered partially configurable products, i.e. one-of-a-kind products that were tailored slightly
beyond the possibilities defined in a configuration model. The modifications often affected only one or two
components. In these cases most of the product individual could be configured and only the customer-specific
modifications had to be designed. Most of the configuration and total cost of the product individual were determined
by the configurable part of the product. Generally it seems more cost-effective to design rare or unpredictable
variants case by case rather than to try to extend the configuration possibilities to cover all possible requirements. It
is a strategic decision whether a company offers customer-specific design on top of configurable products. Note that
a drawback of partially configurable products is that they are difficult to support with configurators.
Sales-delivery process
A request-for-proposal or an order from the customer initiates the configuration process (Figure 2). The customer
requirements and the configuration may be incrementally defined and refined in the configuration process.
Eventually, the configuration should contain all the information necessary for the next stage in the process, e.g. for
manufacturing and installing the product individual. The configuration can be seen as an order to the next stage in
the sales-delivery process.
The process may contain several stages, such as sales and engineering configuration. In the sales stage the product
individual may be specified in terms of the abstract functions or modules that satisfy the customer requirements,
which results in a sales specification or sales order (called configuration 1). However, in most companies the product
individual was specified in technical terms even in the sales phase even though some customers defined their needs
as functions. This means that the sales configurer had to map the functions to technical concepts, usually without
systematic guidelines. In the engineering stage the output of the sales stage is used together with the configuration
model for the engineering stage to produce a more detailed and concrete definition of the product individual
(configuration 2) that can be used as a specification for assembly or manufacturing.
A configuration process in a company may not contain all these stages. For example, the whole configuration task
may be carried out in the sales or engineering stage. When products are relatively simple and easy to configure, sales
configuration can be the only configuration phase. The sales configuration task may sometimes even be carried out
Customer
Requirements
Configuration
model 1
Product
individual
Sales
configuration
Configuration 1
Specification
Configuration
model 2
Assembly
and/or
installation
directly by the customer, for example via Internet
when configuring PCs. However, some complex
products require a product expert to configure
them. In some companies, when sales persons did
not have the required expertise, product experts
were sent to support the local sales organization.
This creates significant costs and is suitable only
for expensive, high-margin products. A configurator enables a company to re-engineer the salesdelivery process of a fully configurable product
and to transfer configuring even a complex product
to the sales person or in some cases to the
customer, without the help of a product expert.
The configuration process typically included at
least finding out the customer requirements,
component selection, determination of the price of
the product individual, preparing the configuration
dependent parts of the proposal, and checking the
Component
Component
completeness and consistency of the configuration.
manufacturing
orders
and/or stock
With some products it was necessary to determine
parameter values for parametric components. The
complexity of the configuration process increased
when layout or component connection design was
Figure 3. A typical sales-delivery process
required. In some cases it was necessary to offer
without configurator support
several alternative configurations for the customer
to choose from. In addition to configuring new
product individuals, some companies needed to modernize, i.e. reconfigure existing product individuals. This task
may be very complex because it bridges the gap between temporally different versions of configuration models. A
modular product architecture facilitates easier modernization of a product individual.
Engineering
configuration
Configuration 2
Description of
product individual
Components
A few companies had good experiences on separate well-defined processes for configurable and partially configurable products. This allowed fluent delivery of configurable products while forcing participants of the process for
partially configurable products to consider all the necessary issues. The processes for partially configurable products
typically included a bid or sales review by a multidisciplinary team including designers, sales persons and production engineers to assess feasibility of the product and estimate the amount of work, time and cost incurred.
Problems in the configuration processes
In many cases 50-80%, sometimes even 100%, of sales specifications were incomplete, that is, some necessary
information was missing. Many sales specifications were also invalid, i.e. they contained errors. According to one
company, 20-25% of the effort of the unit responsible for engineering configuration went into completing incomplete
specifications or correcting invalid ones. The cost of correcting errors in a configuration increases dramatically the
further in the sales-delivery process they are noticed. However, rather few configuration errors reached production or
later phases of the process.
Lead times in the configuration process were long. Plenty of time was used in the configuration process compared to
the actual amount of work. The manufacturing and logistics processes have often been rationalized and automated,
and function efficiently. There is usually more room for improvement in the configuration process than in the
production process.
One of the most common problems of the sales-delivery process (Figure 3) was that incomplete and invalid
configurations caused iteration in the sales-delivery process. It was quite typical that several contacts between the
sales and engineering configurers, the sales configurers and the customers, and even between engineering configurers and the customers were required. Another form of iteration took place when the sales persons did not have the
expertise to configure the product individual on basis of the available, possibly incomplete, configuration model. In
these cases a product expert, typically in a different location, configured the product individual. It often took several
iterations of the loop from the customer to the sales person and further to the engineering configurer and back. This
was necessary to refine the initially vague customer requirements to accurate enough level to fix the final configuration. An iterative process is very error prone due to greater possibility of misunderstandings and frequent changes.
Some of these problems can be attributed to generally poorly documented configuration models. The models were
incomplete, unsystematic, difficult to understand, and relatively often contained outdated information. Order forms
high
One-of-a-kind
product
Price
low
Configurable
product
Mass
product
low
either did not exist or were not systematically used.
Therefore the configurers may not have known the
possibilities of the product and the rules for a complete
and valid configuration. In addition, product policy
seemed to be either loosely defined or it was not
followed. In these cases, the introduction of a product
configurator faces challenges in systematizing both
configuration models and processes.
high
Most of the products were only partially configurable.
There was a strong tradition of one-of-a-kind
production which led to excessive design. Sometimes
changes thought to be minor modifications cost much
Figure 4. Configurable products compared to
more than anticipated. For example, a company
fixed products and one-of-a-kind products
estimated a cost of $11,000 for customer specific
changes that actually cost $29,000. Of the design and management work, 18% was considered “normal”, 65% was
needed for changes and 17% was spent correcting errors caused by the changes.
4.
Adaptability to
customer
requirements
Transition to configurable products
Several companies that had previously delivered one-of-a-kind products had migrated to configurable products to
reduce the lead-times in the sales-delivery process and to re-use the product knowledge (Figure 4). In making this
transition the main effort goes to pre-designing and systemizing the products from design-for-configuration point of
view. The transition may require considerable investments in product development, modularizing the products and
systemizing the product knowledge that often only exists in the designers’ minds. In addition, the sales force must
learn not to offer unnecessary changes to the product that would result in customer-specific design. This cultural
change in the process may be difficult to accept.
The investment in systemizing one-of-a-kind products to be configurable is profitable only if the number of
delivered product individuals is high enough and there is enough similarity between them to allow significant re-use
of the product knowledge in the configuration models. In addition, the product should not be very complex, such as a
plant, nor can the pace of changes to the configuration model be too high. Otherwise managing the configuration
model may become too difficult.
Some companies that had previously offered fixed products had also moved to configurable products (Figure 4). This
has usually addressed the need for a broader product line than can be efficiently managed as fixed products.
However, it is generally not feasible to change very low-priced, really high-volume commodity products to
configurable products. When migrating from fixed products to configurable products, the types of functions that the
customers require may be better known than in case of one-of-a-kind products. Designing a configurable product
may still require considerable effort. Introducing a configuration process to a sales-delivery process that has
previously operated with fixed products can cause problems. More specification work than previously must be done
by the sales persons and in the engineering configuration. All the processes in the company and the supporting
systems must be altered to handle customer-specific variants of products. Changes in processes, information systems,
flexibility of manufacturing and culture may be difficult and expensive.
5.
Configurators
Product configurators have become important support tools for configuration tasks. If used successfully, the benefits
in lead times and quality can be significant. However, there are problems related to their use. The main problem is
related to the long-term management of product knowledge. Many configurator projects have failed due to
difficulties of maintaining configuration models up-to-date. To avoid these problems, a relatively high degree of
maturity in the change processes of the company is necessary. In addition, there is no generally accepted way of
representing a configuration model. Matching a configurator to the products and processes of a company is difficult
but important for a successful application.
A configurator should make it possible to model the configuration model directly with minimal programming effort.
This is a pre-requisite for successful long-term management of product configurators, but it may not be enough. In
addition, a configurator should have the following properties:
• an intuitive, understandable, visual modeling environment;
• a structured, probably object-oriented modeling method; and
• means for modeling the evolution of products, components and their interdependencies.
The first property aims to ensure that the configuration models can be managed by the product developers, product
managers and other people in the company who understand the products best. The second property increases the
understandability of the models and consequently helps managing the model by dividing it into relatively independent packages of knowledge whose interactions are well-defined. The third property adds to the modeling method
some aspects of configuration management[4] and product data management[5]. These may include versions of
products, components and rules on components, as well as effectivity intervals and rules on the different versions. In
our view, there is a clear need for better support for long-term management of configuration models as the pace of
changes to products is increasing.
A considerable problem related to the management of configuration models has also been the effective distribution
of the configurator to the entire sales force. It may be necessary to support both automated and manual configuration
tasks if all retailers (for example, those with low sales volumes) are not willing to acquire or use a configurator. This
problem is worse if a retailer sells products from several manufacturers and should operate several configurators. A
solution to this problem seems to be to centralize the support for configuration tasks and to implement a webbrowser-interface to the centralized configurator. This approach may have problems with security in the Internet but
these can probably be solved with appropriate security technology. More importantly, integration to the information
systems of the retailer can be difficult.
6.
Conclusions
Product configuration is an important research issue and not limited to research on product configurators. A broader
view is very important for companies utilizing configurators. The developers of configurators can also benefit from
such research.
Successful utilization of product configurators necessitates understanding and often re-engineering the sales-delivery
process as well as designing products for configurability. Neglecting these issues in a configurator project often leads
to major difficulties. Long-term management of configuration models should be a major concern. A good match
between the needs of a company and configurator functionality also makes long-term management easier. Some
aspects of practical processes are difficult to support with configurators. These include partially configurable
products, reconfiguration and several stages of the configuration process.
Further research is needed to analyze in-depth the conditions under which a company can benefit from migration to
configurable products and using a configurator. The main factors seem to be the degree to which product knowledge
in configuration models can be re-used, the complexity of the resulting configuration model and the pace of changes
to the configuration models. This issue is being addressed in one of our current research projects.
Acknowledgements
We thank Hannu Peltonen, Asko Martio, and Antti Vainonen for reviewing this paper and for their excellent
comments on how to improve it. This work has been funded by the Technology Development Centre of Finland,
Helsinki Graduate School of Computer Science and Engineering, and the Academy of Finland (project 29655).
References
1. Faltings B. Freuder E. (1996) Configuration—Papers from the 1996 AAAI Fall Symposium. AAAI Press
Technical Report FS-96-03.
2. Tiihonen, J. and Soininen, T. and Männistö, T. and Sulonen, R. (1996). State of the practice in product configuration - a survey of 10 cases in the Finnish industry. Knowledge Intensive CAD, Volume 1, T. Tomiyama, M. Mäntylä,
S. Finger, editors. Chapman & Hall
3. Ulrich, K. and Eppinger, S. (1995). Product Design and Development. McGraw-Hill, Inc.
4. Buckley, F. (1993). Configuration Management. Hardware, Software, and Firmware. IEEE Press.
5. CIMdata, Inc (1997). PDM Buyer’s Guide: Product Data Management Systems for Improving Processes and
Products, Sixth Edition, Volumes 1 and 2.
Download